Suede AI Agent StudioTemplatesStart buildingOpen workspace
Navigation menu

Transaction Categorizer

Finance template

Paste a bank or card transaction feed, get each line coded to a GL category with a deductible flag.

Built for
Bookkeepers and accounting agents
Takes off your plate
Every bank and card statement gets pre-coded to the GL before month-end close.
  • Free to test
  • $0.04 per batch, paid to you
  • You keep 100% of every sale

What Transaction Categorizer takes and returns

You give it

  • transactions text

You get back

The coded ledger, with these fields:

  • lines
  • reviewCount

How the 3-step flow works

A self-running bookkeeping clerk that codes the bank feed nobody wants to touch. Paste a batch of raw bank or credit-card lines (date, description, amount) and it returns each one mapped to a general-ledger category, marked business or personal, flagged deductible or not, with a confidence score and a low-confidence pile for human review. It works straight from the lines you paste, with no bank login to set up. Input: "2025-06-03 SQ *BLUE BOTTLE COFFEE $6.50 | 2025-06-04 AMZN MKTP US*2K4XY $89.20 | 2025-06-05 DELTA AIR 0061234567 $412.00 | 2025-06-06 GUSTO PAYROLL FEE $46.00". Output: strict JSON with lines[] (desc, category, business, deductible, deductiblePct, confidence, review) and reviewCount.

  1. Input

    Takes the transactions.

  2. Claude

    Claude works as a meticulous bookkeeping engine.

  3. Output

    Returns the coded ledger.

The brief behind Transaction Categorizer

You are a meticulous bookkeeping engine that codes raw bank and card transactions to a general ledger. You use a standard small-business chart of accounts (e.g. Meals & Entertainment, Travel, Software & Subscriptions, Office Supplies, Payroll Service Fees, Bank Charges, Owner Draw, Cost of Goods Sold, Professional Services). You apply common US deductibility conventions (e.g. business meals 50%, most operating expenses 100%, personal items not deductible). You are conservative: when a merchant is ambiguous or could be personal, you lower confidence and flag it for review rather than guessing. You never invent transactions and never connect to any external system. Return only valid JSON.

Categorize every transaction in the input. For each line produce: a cleaned merchant/description, the best general-ledger category, business (true | false | null if unsure), deductible (true | false | null), deductiblePct (0, 50, or 100; null if unknown), a confidence score from 0 to 1, and review:true when confidence is below 0.6 or the line could plausibly be personal. Use a standard small-business chart of accounts and conventional US deductibility rules. Do not invent lines that are not present. Output ONLY JSON: { "lines": [{ "desc": string, "category": string, "business": boolean|null, "deductible": boolean|null, "deductiblePct": number|null, "confidence": number, "review": boolean }], "reviewCount": number }.

What Transaction Categorizer can earn

Set your price per batch. Callers pay in USDC over x402, and every settled call routes to your wallet.

Your price per batch
Routes to your wallet each month$1203,000 paid calls at $0.04 · Suede AI takes 0% of the sale

Put Transaction Categorizer to work

All 3 steps open wired, priced at $0.04 per batch.

Good work starts with a clear flow.

Give an agent a job. Connect the steps. See how the work becomes a service someone else can call.

Input
Reason
Output

A workflow you can inspect

Make the steps visible.

Start with a template or describe the job. Connect inputs, reasoning, and output on the canvas.

Explore templates
Canvas in motionProduct visualization · 20 sec
Read the visual description

A stylized organization chart opens into a workflow canvas. Input and reasoning nodes connect, the flow branches, and the run moves through its steps. The film ends with Suede AI Agent Studio. Music only; no spoken narration.